An Effective Feature Extraction Method Used in Breath Analysis

Abstract

It has been reported that human breath could represent some kinds of diseases. By analyzing the components of breath odor, it is easy to detect the diseases the subjects infected. The accuracy of breath analysis depends greatly on what feature are extracted from the response curve of breath analysis system. In this paper, we proposed an effective feature extraction method based on curve fitting for breath analysis, where breath odor were captured and processed by a self-designed breath analysis system. Two parametric analytic models were used to fit the ascending and descending part of the sensor signals respectively, and the set of best-fitting parameters were taken as features. This process is fast, robust, and with less fitting error than other fitting models. Experimental results showed that the features extracted by our method can significantly enhance the performance of subsequent classification algorithms.